An improved KK-means clustering algorithm for fish image segmentation

Fish contour extraction from images is the foundation of many fish image applications such as disease early warning and diagnostics, animal behavior, aquatic product processing, etc. In order to improve the accuracy and stability of fish image segmentation, we propose a new fish images segmentation method which is the combination of the KK-means clustering segmentation algorithm and mathematical morphology. Firstly, the traditional KK-means clustering segmentation algorithm has been improved for fish images. The best number of clusters is determined by the number of gray histogram peaks, and the cluster centers data is filtered by comparing the mean with the threshold decided by Otsu. Secondly, the opening and closing operations of mathematical morphology are used to get the contour of the fish body. The experimental results show that the algorithm realized the separation between the fish image and the background in the condition of complex backgrounds. Compared with Otsu and other segmentation algorithms, our algorithm is more accurate and stable.